 Hello and welcome back to our NVIDIA GTC cover. This is theCUBE, I'm John Furrier with my co's Dave Vellante. We're here, Dave, great to see you. We're here for the kind of the day two, but it's really day one, but all the sessions started but it's really day two. Jensen just did his private briefing with the press and the analysts here with ZS Carvella, ZK Research. Always great to have this analyst angle. Guys, thanks for joining me. Yeah, you're welcome, John. I mean, thanks for having us here. I mean, this is crazy. I mean, this is, talk about Fire Marshall Full. I mean, this place is packed, but, you know, Jensen did a great job yesterday at the keynote, sort of making the case for big honkin' chips. And then we just sat, the three of us just sat through a masterclass. Jensen spoke for an hour, like straight for an hour and then took Q&A from analysts for an hour. Every hand in the room was up, you got a question in. You know, I wasn't able to get my question in, but so good job there, sitting right up front. But, you know, it just strikes me that they're playing in every part of the market, AI everywhere. And my takeaway from listening to yesterday and today is, he's basically saying the same infrastructure is going to do both training and inference. Now, there's definitely a market. He talked about chip lifts a little bit, very little bit, because people have been trying to deposition an NVIDIA's monolithic approach. But there's real benefits to monolithic. And he talked about that, that every part talks to the memory, every part talks to each other. And we had to create a new communication system to be able to do that. So, I think they are taking the high end of the market and there's really nobody that, in my opinion, is going to touch them. Plus, they're also doing chip lift stuff with media tech. The other thing that I wanted to mention is, yesterday we heard about ANSYS, Synopsys, Cadence. No surprise, this rally is broadening. You know, and those companies are getting a lot of AI love. So, it's just AI everywhere. Well, we're here in the run and gun cube, guys. We're one camera, very tight footprint. We've got our cameras. I want to thank Noah and the team back in Palo Alto, making it all happen. We've got Howie Shoe doing a founder AI tracks at the studio. Obviously, it's AI's innovators, our big theme. ZS, we chatted yesterday to kick off the event outside. We did a great remote out front with the cube. We just heard Jensen kind of go kind of deeper under the covers, more personable, a little bit more expansive on his one on one with the ANS. We were just in that room. We had a chance to sit with him and get that vibe. What was your big takeaway that was different than the keynote? How did you share your thoughts? You got a question, congratulations. Yeah, well, he's super excited about it. And I think to follow up on what Dave's points were, was when you think about where NVIDIA came from as a chip company, there could have been a lot of limitations along the way the network's not fast enough, memory doesn't get transmitted fast enough, that could have held them back. And it didn't because they solved it, right? And in fact, we talked a little bit about NVLink yesterday and NVSwitch, where I think that's, I said to Jensen, I think it's the most underappreciated innovation from that company, because it solved the network problem for interchip communications. And I think when you, you're right, they're not a big honk in chips is, I don't know, are they chips, or are they systems, are they motherboards, are they servers, right? It's all those things together. In fact, when they showed the Blackwell video, starting with the chip, then going to two chips, then going to a quad chip, to a motherboard, to a server, to a system, to a cluster of systems, right? That's really, yeah, and they build all of that together. And so are they vertically integrated, sure. But when you think of, up until this point, what's the best performing enterprise applications ever been? It's Oracle, right? And it's Oracle because it's a fully integrated vertical system. And when performance matters, nothing outperforms engineered systems. And so I think they've taken that and went into the nth degree. And the thing that impresses me about NVIDIA is whatever limitations there are along the way, they fix them, right? And whether they have to reinvent networking or memory, they do that. And I think that's a, it's not a company that can do that. You know, Jay, we're going to be sitting down tomorrow with Broadcom's financial analyst meetings here in San Jose. And we had Charlie Quas on theCUBE at MWC and you hear Jensen talk. And one of the things that strikes me about this generation, these leaders in the chips, and then switches, actually they have chips and switches, both of those companies, by the only companies that have both switches and chips. Unique, Broadcom and NVIDIA. But their CEOs are computer science guys. So when they explain concepts, it's refreshing because they have two degrees of computer science to hear the talk. And he's kind of simplifying computer science principles. And you can hear, you can see how much of their lead that they have. They've made big bets, really quietly made big bets. They put ray tracing and all the stuff. They have a tensor core and all the millions of PCs. Three years and 25,000 employees working on Blackwell. Big bets in zero billion markets that become huge. Everyone thinks they see the success but they don't see the failures. It just strikes me that they got it right and these bets are paying off and they have a huge lead. These systems are the way they're engineered. It's a data center basically as a service. They're selling as parts he said. So really interesting that the moat that they have Dave is really their lead. What do you guys think about that? What's your reaction to that whole size of the lead that they have in these markets? And what's real and what may not be as tight as a drum as you might think? Well the line that was an awesome line when he said we are market makers not share takers. So basically he's saying we're creating markets. Now, believe me, they're taking share. They're taking share from x86. That's like Broadcom saying they don't chase S-curves but they're on one. But exactly, by accident, by design. But so I think that their moat is substantial. I mean, I think when, you know, we've heard leading up to GTC a lot of talk about monolithic and almost used as a pejorative. To me, monolithic's not a pejorative because you need, as Jensen said, you need bigger and bigger and bigger chips. And his point is we share memory across all these components. And so when you even talked about the cabling, you talked about all the potential constraints. Cabling was one potential. They had to basically invent some new switching capabilities. So I think they're far ahead in packaging and obviously in chip design. I think they've got yield advantages with TSM. I mean, on and on and on. And just to hear Jensen talk, he doesn't talk like a chip manufacturer. He talks like a system, a design person, system thinking. And he really talks about the future, the future of everything. And it's extremely impressive. And I think it really underscores their lead. Well, the other part of the lead though is just the size of the developer community, right? When even if somebody else were to come up and build the software needed and the developer tools needed, how long would it take you to build that developer ecosystem? I think he said in the keynote yesterday what was a 1.3 million robotic developers alone, right? So from a mode perspective, even if you were at par on the tech, which competitors aren't, it would take you, what, decade or so to go accumulate all those developers? And so that's always been the heart of this company, I find, and really the most misunderstood part of NVIDIA is they've done the things they need to do to make it easy for developers. And other chip companies didn't, for whatever reason, see that and NVIDIA is really the one winning the war now. Actually, I loved your question about accelerated computing and how you tied it to the network. But then he kind of went out in a tangent and it got my attention, Dave, because Zia's got that question in, but it sparked something in Jensen's head that he answered the question about IT, Swiss traffic, and AI. But he actually went on this tangent where he talked about the AI acceleration to industries. He meant, and he kind of was not bragging. It's not bragging if it's true, by the way. So he wasn't really bragging. He was saying, yeah, we have a lead and no one's really chasing us. But I think we're going to see a new dynamic with this accelerated computing that we've seen before, but not the same. Remember during the pandemic, there was this concept that was in the financial world called pull forward revenue. Yeah, the whole pull forward opportunities that because of the pandemic created new use cases. And some thought they were mirages, some thought they were not sustainable, but they had a huge financial impact. So this pull forward impact that the pandemic had, AI is happening. So what he's basically saying today was, I heard him say was, we're pulling forward from the future, accelerating use cases with our AI, rather than net new invention, we're going to have incremental improvement and bring forward some of these advances faster. And he named multiple use cases almost in every industry. We've been kind of saying that Dave, like AI is great in the industry. So I think this pull forward concept is going to be huge. I think it's going to have a material impact, not only on financial, but society. So I think this accelerated computing, accelerated AI is going to absolutely have an impact. And again, you're starting to see the early signs of it. And now you get the systems and the hardware in place. The chip side, I love that big chip, build a big chip because there's efficiencies with energy and chips. So there's huge advantages to the architecture, the business model, and then ultimately the impact. I think the concern, if I may, about the pull forward and use the pandemic, and I think this is a really good example, what happened in the pandemic is we got this sort of seesaw slingshot effect where everybody ramped up and then everybody had to ramp down because everything changed. And so people are obviously concerned about that pull forward effect. I mean, if you want a Blackwell, you got to sign up for Blackwell now or you go back to the end of the line. So of course, who's signing up? AWS, Microsoft, Google, Oracle, Dell, et cetera, they're all saying, yeah, give us the chips. And so you may see that type of slingshot effect, but look, it's better to be early and overbuilt than to be underbuilt and not be there to take advantage of the opportunity. And the other key status, he talked about 25,000 engineers, you just said over three years, $10 billion to build this thing. So at 50,000 bucks a piece, you're talking about a couple of years before he gets that payback. And then the other question that came up, somebody asked about software, like what's your software opportunity? He said, well, imagine we got a million GPUs out there at $4,500 a piece per GPU per year, you're talking about a four or $5 billion opportunity per year. They're going to make a lot of money. The thing I want to get you guys' thoughts on is that he mentioned a couple of things. We're going to manufacture AI and LLMs at scale and create LLMs, going to be in the LLM creation business as well as working with others. He mentioned someone's got to pay for this operating system. That's, he's hinting that they're the operating system for the future. But he started the whole thing with this new industrial revolution. And I find this interesting. I think NVIDIA is making a land grab for the token market, Dave. The tokens were, came up multiple times on the keynote. He drilled on it at the beginning of his keynote here in this new industrial revolution pitch, where he said, the old world is prerecorded. Like we think prerecorded videos, you know, you get the news, it's prerecorded, everything's prerecorded, you got to fetch it versus generative. Seeds of innovation. Data is seeds, the new seeds. Retrieval versus generation. So I think this whole token economics redo, remember the old crypto days, Dave? ICOs and token economics. I think you're going to see a token currency. And I think they're smart to build the system, to handle LLMs at scale, because he mentioned the word embeddings. He wants that embedding business. I think this is a big strategic move for NVIDIA to go after the tokens, because they can control that context window, they can control the reg and ultimately the LLMs. And if they have the factory, they got to have the embeds, Dave. The embeds is going to be the core currency for those tokens. So, did you hear that? What's your reaction to that? Yeah, no, I think ultimately the operating system analogy is the right one. In fact, during his keynote, he talked about Omniverse, being the operating system for digital twins. And he talked about LLMs, what they're doing there is being the, really effectively the operating system for content generation. So when you think of, once you're the operating system, then the ecosystems that get built on you become massive and create this pull through effect for me, right? So, I did like the comparison to operating system because I think it is an appropriate one. But clearly as we move into this world where more and more stuff gets generated on the fly, I mean, that's awfully hard for anybody else to do without owning the stack. What I like about Chesham, he obviously says some things over and over again like AI factories, and if you buy more, you save more. But he also introduces new content every time I see him talk. And I'll give you some examples here. He talked about the world of generated content is here. And he talked about how NVIDIA creates a lot of content formally through graphics and now they're generating tons of content. Yeah, he talked about that almost as being the renaissance for the GPU, right? Cause they started off as a company that created graphics and now they're creating all multimodal content. And he said Blackwell is built for this gen AI moment. And then he went through this thing of talking about simulation versus animation and how you do training and learning and generation. And then you have to ground the AI in physics and that's what Omniverse does. And so my point is, when Jensen talks, he talks with vision. When his competitors talk, they talk about, well, GPUs are really expensive and we're going to lower the cost of GPUs and we're going to compete. He talks about how he's changing the world. He said none of the stuff that I talk about was ever here before, gen AI. And this is all new. So his vision is really crisp and quite forward thinking. One of the pieces he hinted at at the end that I wish he'd have spent more time on was the sustainability impact of what NVIDIA is doing because that whenever I read media articles about what to expect, NVIDIA gets beat up a lot because Blackwell uses a lot of power. But if you think about the workloads they're running, now it's a weird comparison because you wouldn't run those workloads without Blackwell but now that you are, Blackwell uses a lot more power. But you can do things a lot more power efficiently. And so when he talked about embedding GPUs in AI into cellular radios, right? Instead of just sort of upping the power all the time, it can be a lot smarter about how it does that. So are they using more power? Sure, but we're talking about this on the way over that the power per transmitted packet goes way down and that's the calculation to do. And I think that's an opportunity for NVIDIA. They really, he touches on a little bit but given how important sustainability is, it's something you can drill down a lot more on. He just drops a lot of nuggets because he's kind of off script so he kind of slips in because we're reading every word and connecting the dots. SmartNix and all nodes, you catch that one? Oh yeah. Okay, that was good. East-West, Northwest encryption. You brought that up in your question. But the SmartNix and all the nodes is a decoupling and that really points to their success with MV Switch Day because if you look at their success with the new chip is unbelievable how they had that one chip that handles all the GPUs acting as one and then now that the Switch chip is by itself that's going to be a major innovation. And again, this is what we just talked about at MWC. The chips working together, IO, low energy, good yield. This is a huge issue because basically when people talk about these monolithic chips what they're talking about is you have a huge SRAM that's shared amongst all these components and those are synchronous connections whereas again, these alternative designs people can't get yields out of what Jensen is doing and somehow Jensen and NVIDIA have figured out how to get yields. So what they do is they build chiplets which have their place but they're asynchronous connections and by comparison they're much, much slower. The other mic drop moment I thought was he said NVIDIA is the world's largest quantum computing company that doesn't build quantum computers. And then he went off and started talking about- Kind of disinquantum. Yeah, he kind of was. He was saying we are quantum without the computer. Basically AI is going to solve a lot of these problems. Except for use case, except for specific use case things. Which was encryption, quantum encryption. He said AI is not going to solve that but he gave several examples which could be boiled down to pretty simple problem solving that he said AI was going to solve. He did shoot an arrow across the ethernet bow and the link can't go down. Super high quality. What do you think about that? Energy is low. I said this morning to Kevin at Dearly Runs Networking. They announced about the holy war between ethernet and infinite band and he said there's no holy war. But then he went on to say that you cannot use ethernet and AI. It sucks. So there is a holy war going on here. I know all the network vendors are part of the Altered Ethernet Forum but as you and I have talked about before I've seen this movie before where everyone wants ethernet to replace infinite band. But frankly if you look at the relative cost of networking within the AI cluster it's not all that much, right? And so the risk reward on replacing it with something that may or may not work to me isn't all that high. So I do think once you're outside the cluster you're going to go ethernet and out through ethernet might be good for that. And so it'll, ethernet will lead away at infinite band kind of at the margins but infinite band's here for a while. There's markets for both. And this is where when you sit down with Charlie Kawas and people like that who have deep technical expertise and he talks about open scale and power it's like okay this guy makes a lot of sense. And so I get, I think the takeaway John is there's markets for both. Yeah broad comment and video will be the dominant. All right guys last question real quick as we wrap this analyst angle up. This is a, he said the next decade don't be asleep for this decade. We're two years in for the next 10 years. Jensen said that today. We've been saying that in theCUBE. We wish we were 25 again and since a great run we're absolutely going to be paying attention and living this next 10 years. What's the impact of this NVIDIA GTC this year? I mean outside we're getting in the weeds but mainstream and people see the stock price. They know GPUs. This is a computer industry that's being redefined and he's actually saying it's the new industrial revolution obviously accelerated computing there at the head of the pack. What's the impact of this GTC? Dave we'll start with you. That's going to, is it a market history? What's your takeaway of the impact of this year? I think it is a market history. I think people are concerned. They're like okay when's NVIDIA going to run out of gas? What's the competition look like? What does that mean for NVIDIA's stock price? I think we are in the early days of a data center super cycle and I think it is going to morph the way that mobile and cloud changed our lives. I think it's going to be completely radical. He said within eight years and I agree with him we're going to have functional robots. So we're going to be talking to robots and they're going to be doing these menial tasks around our homes and it's going to change the way. I agree. It's going to change the way. You know we have to figure out other ways to get our exercise but it's going to change the way in which people live and work and we're just in the early innings of a 10 year data center super cycle. No robots on cube though, right? I don't know, we'll see. We'll see. I agree with you and I think with AI. We've been talking about AI for a long time so if you think of the typical hockey stick and option of technology, I think the AI blade's been a little longer than cloud and things like that but I think the shaft is going to be a lot steeper and a lot longer. And I do think we are in the very early innings here. One of the execs this morning in the Q and A said he thinks we're at first pitch. Not even first inning, right? So from that standpoint, NVIDIA's got a tremendous amount of runway. I think one of the other interesting things to think about just from a capital markets perspective is we're going to have some big tailwinds with AI. Who are the other companies that are going to take advantage of this but some may have to burn themselves down to be able to be reborn in this era but I think there's some obvious ones like Pure Storage and Arista maybe but I think the rest of that ecosystem is yet to be written. Reinforced learning, big theme, tokens, grounded in truth. This is the AI future where the cube is here on the ground running and gunning with a single camera. Nothing's going to stop us from getting the stories. We do whatever it takes to get that story out to you and thanks for watching this analyst angle on the cube. We're in systems revolution. Next 10 years is going to be a lot of fun, a lot of change, a lot of great opportunities to make a difference. We'll be there with you. Thanks for watching.